A Supercomputer-Based Modeling System for Short-Term Prediction of Urban Surface Air Quality

Abstract

This paper proposes a mathematical model and an effective supercomputer-based numerical method for short-term prediction of extreme meteorological conditions and atmospheric air quality over limited stretches of land encompassing large population centers. The mathematical model includes a pollutant transport model with a reduced chemical mechanism and a non-hydrostatic mesoscale meteorological model with a modern moisture microphysics parametrization scheme. The numerical method relies on the use of the finite volume method and semi-implicit difference schemes of the second order of approximation, which are solved using the TDMA method with a linear dependence of the number of arithmetic operations on the size of the grid. This property of the numerical method ensures high efficiency when parallelized: not less than 70% when using up to 256 computing cores with a horizontal grid size of 0.5–1.0 km. Development of parallel programs was carried out using the Message Passing Interface parallel programming technology, two-dimensional decomposition of the grid area along horizontal (west to east and south to north) directions, and introduction of additional fictitious grid nodes along the perimeter of the decomposition subdomains